Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Mukesh Ojha <quic_mojha@quicinc.com>
To:     <linux-kernel@vger.kernel.org>
CC:     <gregkh@linuxfoundation.org>, <tglx@linutronix.de>,
        <sboyd@kernel.org>, <rafael@kernel.org>,
        <johannes@sipsolutions.net>, <keescook@chromium.org>,
        Mukesh Ojha <quic_mojha@quicinc.com>
Subject: [PATCH] devcoredump : Serialize devcd_del work
Date:   Mon, 2 May 2022 21:04:31 +0530
Message-ID: <1651505671-1506-1-git-send-email-quic_mojha@quicinc.com>
MIME-Version: 1.0
Content-Type: text/plain
Precedence: bulk

In following scenario(diagram), when one thread X running dev_coredumpm()
adds devcd device to the framework which sends uevent notification to
userspace and another thread Y reads this uevent and call to
devcd_data_write() which eventually try to delete the queued timer that
is not initialized/queued yet.

So, debug object reports some warning and in the meantime, timer is
initialized and queued from X path. and from Y path, it gets reinitialized
again and timer->entry.pprev=NULL and try_to_grab_pending() stucks.

To fix this, introduce mutex and a boolean flag to serialize the behaviour.

 	cpu0(X)			                cpu1(Y)

    dev_coredump() uevent sent to user space
    device_add()  ======================> user space process Y reads the
                                          uevents writes to devcd fd
                                          which results into writes to

                                         devcd_data_write()
                                           mod_delayed_work()
                                             try_to_grab_pending()
                                               del_timer()
                                                 debug_assert_init()
   INIT_DELAYED_WORK()
   schedule_delayed_work()
                                                   debug_object_fixup()
                                                     timer_fixup_assert_init()
                                                       timer_setup()
                                                         do_init_timer()
                                                       /*
                                                        Above call reinitializes
                                                        the timer to
                                                        timer->entry.pprev=NULL
                                                        and this will be checked
                                                        later in timer_pending() call.
                                                       */
                                                 timer_pending()
                                                  !hlist_unhashed_lockless(&timer->entry)
                                                    !h->pprev
                                                /*
                                                  del_timer() checks h->pprev and finds
                                                  it to be NULL due to which
                                                  try_to_grab_pending() stucks.
                                                */

Link: https://lore.kernel.org/lkml/2e1f81e2-428c-f11f-ce92-eb11048cb271@quicinc.com/
Signed-off-by: Mukesh Ojha <quic_mojha@quicinc.com>
---
v3->v4:
 - flg variable renamed to delete_work.

v2->v3:
 Addressed comments from gregkh
 - Wrapped the commit text and corrected the alignment.
 - Described the reason to introduce new variables.
 - Restored the blank line.
 - rename the del_wk_queued to flg.
 Addressed comments from tglx
 - Added a comment which explains the race which looks obvious however
   would not occur between disabled_store and devcd_del work.


v1->v2:
 - Added del_wk_queued flag to serialize the race between devcd_data_write()
   and disabled_store() => devcd_free().
 drivers/base/devcoredump.c | 83 ++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 81 insertions(+), 2 deletions(-)

diff --git a/drivers/base/devcoredump.c b/drivers/base/devcoredump.c
index f4d794d..1c06781 100644
--- a/drivers/base/devcoredump.c
+++ b/drivers/base/devcoredump.c
@@ -25,6 +25,47 @@ struct devcd_entry {
 	struct device devcd_dev;
 	void *data;
 	size_t datalen;
+	/*
+	 * Here, mutex is required to serialize the calls to del_wk work between
+	 * user/kernel space which happens when devcd is added with device_add()
+	 * and that sends uevent to user space. User space reads the uevents,
+	 * and calls to devcd_data_write() which try to modify the work which is
+	 * not even initialized/queued from devcoredump.
+	 *
+	 *
+	 *
+	 *        cpu0(X)                                 cpu1(Y)
+	 *
+	 *        dev_coredump() uevent sent to user space
+	 *        device_add()  ======================> user space process Y reads the
+	 *                                              uevents writes to devcd fd
+	 *                                              which results into writes to
+	 *
+	 *                                             devcd_data_write()
+	 *                                               mod_delayed_work()
+	 *                                                 try_to_grab_pending()
+	 *                                                   del_timer()
+	 *                                                     debug_assert_init()
+	 *       INIT_DELAYED_WORK()
+	 *       schedule_delayed_work()
+	 *
+	 *
+	 * Also, mutex alone would not be enough to avoid scheduling of
+	 * del_wk work after it get flush from a call to devcd_free()
+	 * mentioned as below.
+	 *
+	 *	disabled_store()
+	 *        devcd_free()
+	 *          mutex_lock()             devcd_data_write()
+	 *          flush_delayed_work()
+	 *          mutex_unlock()
+	 *                                   mutex_lock()
+	 *                                   mod_delayed_work()
+	 *                                   mutex_unlock()
+	 * So, delete_work flag is required.
+	 */
+	struct mutex mutex;
+	bool delete_work;
 	struct module *owner;
 	ssize_t (*read)(char *buffer, loff_t offset, size_t count,
 			void *data, size_t datalen);
@@ -84,7 +125,12 @@ static ssize_t devcd_data_write(struct file *filp, struct kobject *kobj,
 	struct device *dev = kobj_to_dev(kobj);
 	struct devcd_entry *devcd = dev_to_devcd(dev);
 
-	mod_delayed_work(system_wq, &devcd->del_wk, 0);
+	mutex_lock(&devcd->mutex);
+	if (!devcd->delete_work) {
+		devcd->delete_work = true;
+		mod_delayed_work(system_wq, &devcd->del_wk, 0);
+	}
+	mutex_unlock(&devcd->mutex);
 
 	return count;
 }
@@ -112,7 +158,12 @@ static int devcd_free(struct device *dev, void *data)
 {
 	struct devcd_entry *devcd = dev_to_devcd(dev);
 
+	mutex_lock(&devcd->mutex);
+	if (!devcd->delete_work)
+		devcd->delete_work = true;
+
 	flush_delayed_work(&devcd->del_wk);
+	mutex_unlock(&devcd->mutex);
 	return 0;
 }
 
@@ -122,6 +173,30 @@ static ssize_t disabled_show(struct class *class, struct class_attribute *attr,
 	return sysfs_emit(buf, "%d\n", devcd_disabled);
 }
 
+/*
+ *
+ *	disabled_store()                                	worker()
+ *	 class_for_each_device(&devcd_class,
+ *		NULL, NULL, devcd_free)
+ *         ...
+ *         ...
+ *	   while ((dev = class_dev_iter_next(&iter))
+ *                                                             devcd_del()
+ *                                                               device_del()
+ *                                                                 put_device() <- last reference
+ *             error = fn(dev, data)                           devcd_dev_release()
+ *             devcd_free(dev, data)                           kfree(devcd)
+ *             mutex_lock(&devcd->mutex);
+ *
+ *
+ * In the above diagram, It looks like disabled_store() would be racing with parallely
+ * running devcd_del() and result in memory abort while acquiring devcd->mutex which
+ * is called after kfree of devcd memory  after dropping its last reference with
+ * put_device(). However, this will not happens as fn(dev, data) runs
+ * with its own reference to device via klist_node so it is not its last reference.
+ * so, above situation would not occur.
+ */
+
 static ssize_t disabled_store(struct class *class, struct class_attribute *attr,
 			      const char *buf, size_t count)
 {
@@ -278,13 +353,16 @@ void dev_coredumpm(struct device *dev, struct module *owner,
 	devcd->read = read;
 	devcd->free = free;
 	devcd->failing_dev = get_device(dev);
+	devcd->delete_work = false;
 
+	mutex_init(&devcd->mutex);
 	device_initialize(&devcd->devcd_dev);
 
 	dev_set_name(&devcd->devcd_dev, "devcd%d",
 		     atomic_inc_return(&devcd_count));
 	devcd->devcd_dev.class = &devcd_class;
 
+	mutex_lock(&devcd->mutex);
 	if (device_add(&devcd->devcd_dev))
 		goto put_device;
 
@@ -301,10 +379,11 @@ void dev_coredumpm(struct device *dev, struct module *owner,
 
 	INIT_DELAYED_WORK(&devcd->del_wk, devcd_del);
 	schedule_delayed_work(&devcd->del_wk, DEVCD_TIMEOUT);
-
+	mutex_unlock(&devcd->mutex);
 	return;
  put_device:
 	put_device(&devcd->devcd_dev);
+	mutex_unlock(&devcd->mutex);
  put_module:
 	module_put(owner);
  free:
-- 
2.7.4