[PATCH AUTOSEL 5.15 29/51] kernfs: Separate kernfs_pr_cont_buf and rename_lock.

[Sasha Levin]

From: Hao Luo <haoluo@xxxxxxxxxx>

[ Upstream commit 1a702dc88e150487c9c173a249b3d236498b9183 ]

Previously the protection of kernfs_pr_cont_buf was piggy backed by
rename_lock, which means that pr_cont() needs to be protected under
rename_lock. This can cause potential circular lock dependencies.

If there is an OOM, we have the following call hierarchy:

 -> cpuset_print_current_mems_allowed()
   -> pr_cont_cgroup_name()
     -> pr_cont_kernfs_name()

pr_cont_kernfs_name() will grab rename_lock and call printk. So we have
the following lock dependencies:

 kernfs_rename_lock -> console_sem

Sometimes, printk does a wakeup before releasing console_sem, which has
the dependence chain:

 console_sem -> p->pi_lock -> rq->lock

Now, imagine one wants to read cgroup_name under rq->lock, for example,
printing cgroup_name in a tracepoint in the scheduler code. They will
be holding rq->lock and take rename_lock:

 rq->lock -> kernfs_rename_lock

Now they will deadlock.

A prevention to this circular lock dependency is to separate the
protection of pr_cont_buf from rename_lock. In principle, rename_lock
is to protect the integrity of cgroup name when copying to buf. Once
pr_cont_buf has got its content, rename_lock can be dropped. So it's
safe to drop rename_lock after kernfs_name_locked (and
kernfs_path_from_node_locked) and rely on a dedicated pr_cont_lock
to protect pr_cont_buf.

Acked-by: Tejun Heo <tj@xxxxxxxxxx>
Signed-off-by: Hao Luo <haoluo@xxxxxxxxxx>
Link: https://lore.kernel.org/r/20220516190951.3144144-1-haoluo@xxxxxxxxxx
Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
---
 fs/kernfs/dir.c | 31 +++++++++++++++++++------------
 1 file changed, 19 insertions(+), 12 deletions(-)

diff --git a/fs/kernfs/dir.c b/fs/kernfs/dir.c
index 8e0a1378a4b1..7bf1d5fc2e9c 100644
--- a/fs/kernfs/dir.c
+++ b/fs/kernfs/dir.c
@@ -19,7 +19,15 @@
 
 DECLARE_RWSEM(kernfs_rwsem);
 static DEFINE_SPINLOCK(kernfs_rename_lock);	/* kn->parent and ->name */
-static char kernfs_pr_cont_buf[PATH_MAX];	/* protected by rename_lock */
+/*
+ * Don't use rename_lock to piggy back on pr_cont_buf. We don't want to
+ * call pr_cont() while holding rename_lock. Because sometimes pr_cont()
+ * will perform wakeups when releasing console_sem. Holding rename_lock
+ * will introduce deadlock if the scheduler reads the kernfs_name in the
+ * wakeup path.
+ */
+static DEFINE_SPINLOCK(kernfs_pr_cont_lock);
+static char kernfs_pr_cont_buf[PATH_MAX];	/* protected by pr_cont_lock */
 static DEFINE_SPINLOCK(kernfs_idr_lock);	/* root->ino_idr */
 
 #define rb_to_kn(X) rb_entry((X), struct kernfs_node, rb)
@@ -230,12 +238,12 @@ void pr_cont_kernfs_name(struct kernfs_node *kn)
 {
 	unsigned long flags;
 
-	spin_lock_irqsave(&kernfs_rename_lock, flags);
+	spin_lock_irqsave(&kernfs_pr_cont_lock, flags);
 
-	kernfs_name_locked(kn, kernfs_pr_cont_buf, sizeof(kernfs_pr_cont_buf));
+	kernfs_name(kn, kernfs_pr_cont_buf, sizeof(kernfs_pr_cont_buf));
 	pr_cont("%s", kernfs_pr_cont_buf);
 
-	spin_unlock_irqrestore(&kernfs_rename_lock, flags);
+	spin_unlock_irqrestore(&kernfs_pr_cont_lock, flags);
 }
 
 /**
@@ -249,10 +257,10 @@ void pr_cont_kernfs_path(struct kernfs_node *kn)
 	unsigned long flags;
 	int sz;
 
-	spin_lock_irqsave(&kernfs_rename_lock, flags);
+	spin_lock_irqsave(&kernfs_pr_cont_lock, flags);
 
-	sz = kernfs_path_from_node_locked(kn, NULL, kernfs_pr_cont_buf,
-					  sizeof(kernfs_pr_cont_buf));
+	sz = kernfs_path_from_node(kn, NULL, kernfs_pr_cont_buf,
+				   sizeof(kernfs_pr_cont_buf));
 	if (sz < 0) {
 		pr_cont("(error)");
 		goto out;
@@ -266,7 +274,7 @@ void pr_cont_kernfs_path(struct kernfs_node *kn)
 	pr_cont("%s", kernfs_pr_cont_buf);
 
 out:
-	spin_unlock_irqrestore(&kernfs_rename_lock, flags);
+	spin_unlock_irqrestore(&kernfs_pr_cont_lock, flags);
 }
 
 /**
@@ -822,13 +830,12 @@ static struct kernfs_node *kernfs_walk_ns(struct kernfs_node *parent,
 
 	lockdep_assert_held_read(&kernfs_rwsem);
 
-	/* grab kernfs_rename_lock to piggy back on kernfs_pr_cont_buf */
-	spin_lock_irq(&kernfs_rename_lock);
+	spin_lock_irq(&kernfs_pr_cont_lock);
 
 	len = strlcpy(kernfs_pr_cont_buf, path, sizeof(kernfs_pr_cont_buf));
 
 	if (len >= sizeof(kernfs_pr_cont_buf)) {
-		spin_unlock_irq(&kernfs_rename_lock);
+		spin_unlock_irq(&kernfs_pr_cont_lock);
 		return NULL;
 	}
 
@@ -840,7 +847,7 @@ static struct kernfs_node *kernfs_walk_ns(struct kernfs_node *parent,
 		parent = kernfs_find_ns(parent, name, ns);
 	}
 
-	spin_unlock_irq(&kernfs_rename_lock);
+	spin_unlock_irq(&kernfs_pr_cont_lock);
 
 	return parent;
 }
-- 
2.35.1