mm: shrinkers: fix race condition on debugfs cleanup

When something registers and unregisters many shrinkers, such as:
    for x in $(seq 10000); do unshare -Ui true; done

Sometimes the following error is printed to the kernel log:
    debugfs: Directory '...' with parent 'shrinker' already present!

This occurs since commit badc28d4924b ("mm: shrinkers: fix deadlock in
shrinker debugfs") / v6.2: Since the call to `debugfs_remove_recursive`
was moved outside the `shrinker_rwsem`/`shrinker_mutex` lock, but the call
to `ida_free` stayed inside, a newly registered shrinker can be
re-assigned that ID and attempt to create the debugfs directory before the
directory from the previous shrinker has been removed.

The locking changes in commit f95bdb700bc6 ("mm: vmscan: make global slab
shrink lockless") made the race condition more likely, though it existed
before then.

Commit badc28d4924b ("mm: shrinkers: fix deadlock in shrinker debugfs")
could be reverted since the issue is addressed should no longer occur
since the count and scan operations are lockless since commit 20cd1892fcc3
("mm: shrinkers: make count and scan in shrinker debugfs lockless"). 
However, since this is a contended lock, prefer instead moving `ida_free`
outside the lock to avoid the race.

Link: https://lkml.kernel.org/r/20230503013232.299211-1-joanbrugueram@gmail.com
Fixes: badc28d4924b ("mm: shrinkers: fix deadlock in shrinker debugfs")
Signed-off-by: Joan Bruguera Micó <joanbrugueram@gmail.com>
Cc: Qi Zheng <zhengqi.arch@bytedance.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

3 files changed, 24 insertions, 9 deletions

diff --git a/include/linux/shrinker.h b/include/linux/shrinker.h
index 7bde8e1c228a..224293b2dd06 100644
--- a/include/linux/shrinker.h
+++ b/include/linux/shrinker.h
@@ -107,7 +107,10 @@ extern void synchronize_shrinkers(void);
 
 #ifdef CONFIG_SHRINKER_DEBUG
 extern int shrinker_debugfs_add(struct shrinker *shrinker);
-extern struct dentry *shrinker_debugfs_remove(struct shrinker *shrinker);
+extern struct dentry *shrinker_debugfs_detach(struct shrinker *shrinker,
+					      int *debugfs_id);
+extern void shrinker_debugfs_remove(struct dentry *debugfs_entry,
+				    int debugfs_id);
 extern int __printf(2, 3) shrinker_debugfs_rename(struct shrinker *shrinker,
 						  const char *fmt, ...);
 #else /* CONFIG_SHRINKER_DEBUG */
@@ -115,10 +118,16 @@ static inline int shrinker_debugfs_add(struct shrinker *shrinker)
 {
 	return 0;
 }
-static inline struct dentry *shrinker_debugfs_remove(struct shrinker *shrinker)
+static inline struct dentry *shrinker_debugfs_detach(struct shrinker *shrinker,
+						     int *debugfs_id)
 {
+	*debugfs_id = -1;
 	return NULL;
 }
+static inline void shrinker_debugfs_remove(struct dentry *debugfs_entry,
+					   int debugfs_id)
+{
+}
 static inline __printf(2, 3)
 int shrinker_debugfs_rename(struct shrinker *shrinker, const char *fmt, ...)
 {
diff --git a/mm/shrinker_debug.c b/mm/shrinker_debug.c
index 3f83b10c5031..fe10436d9911 100644
--- a/mm/shrinker_debug.c
+++ b/mm/shrinker_debug.c
@@ -237,7 +237,8 @@ int shrinker_debugfs_rename(struct shrinker *shrinker, const char *fmt, ...)
 }
 EXPORT_SYMBOL(shrinker_debugfs_rename);
 
-struct dentry *shrinker_debugfs_remove(struct shrinker *shrinker)
+struct dentry *shrinker_debugfs_detach(struct shrinker *shrinker,
+				       int *debugfs_id)
 {
 	struct dentry *entry = shrinker->debugfs_entry;
 
@@ -246,14 +247,18 @@ struct dentry *shrinker_debugfs_remove(struct shrinker *shrinker)
 	kfree_const(shrinker->name);
 	shrinker->name = NULL;
 
-	if (entry) {
-		ida_free(&shrinker_debugfs_ida, shrinker->debugfs_id);
-		shrinker->debugfs_entry = NULL;
-	}
+	*debugfs_id = entry ? shrinker->debugfs_id : -1;
+	shrinker->debugfs_entry = NULL;
 
 	return entry;
 }
 
+void shrinker_debugfs_remove(struct dentry *debugfs_entry, int debugfs_id)
+{
+	debugfs_remove_recursive(debugfs_entry);
+	ida_free(&shrinker_debugfs_ida, debugfs_id);
+}
+
 static int __init shrinker_debugfs_init(void)
 {
 	struct shrinker *shrinker;
diff --git a/mm/vmscan.c b/mm/vmscan.c
index d257916f39e5..6d0cd2840cf0 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -805,6 +805,7 @@ EXPORT_SYMBOL(register_shrinker);
 void unregister_shrinker(struct shrinker *shrinker)
 {
 	struct dentry *debugfs_entry;
+	int debugfs_id;
 
 	if (!(shrinker->flags & SHRINKER_REGISTERED))
 		return;
@@ -814,13 +815,13 @@ void unregister_shrinker(struct shrinker *shrinker)
 	shrinker->flags &= ~SHRINKER_REGISTERED;
 	if (shrinker->flags & SHRINKER_MEMCG_AWARE)
 		unregister_memcg_shrinker(shrinker);
-	debugfs_entry = shrinker_debugfs_remove(shrinker);
+	debugfs_entry = shrinker_debugfs_detach(shrinker, &debugfs_id);
 	mutex_unlock(&shrinker_mutex);
 
 	atomic_inc(&shrinker_srcu_generation);
 	synchronize_srcu(&shrinker_srcu);
 
-	debugfs_remove_recursive(debugfs_entry);
+	shrinker_debugfs_remove(debugfs_entry, debugfs_id);
 
 	kfree(shrinker->nr_deferred);
 	shrinker->nr_deferred = NULL;