Re: [PATCH] Fix of dcache race leading to busy inodes on umount

[Kirill Korotaev <dev@sw.ru>]

>>This patch fixes dcache race between shrink_dcache_XXX functions
>>and dput().
> 
> 
> 
>>-void dput(struct dentry *dentry)
>>+static void dput_recursive(struct dentry *dentry, struct dcache_shrinker *ds)
>> {
>>-	if (!dentry)
>>-		return;
>>-
>>-repeat:
>> 	if (atomic_read(&dentry->d_count) == 1)
>> 		might_sleep();
>> 	if (!atomic_dec_and_lock(&dentry->d_count, &dcache_lock))
>> 		return;
>>+	dcache_shrinker_del(ds);
>> 
>>+repeat:
>> 	spin_lock(&dentry->d_lock);
>> 	if (atomic_read(&dentry->d_count)) {
>> 		spin_unlock(&dentry->d_lock);
>>@@ -182,6 +253,7 @@ unhash_it:
>> 
>> kill_it: {
>> 		struct dentry *parent;
>>+		struct dcache_shrinker lds;
>> 
>> 		/* If dentry was on d_lru list
>> 		 * delete it from there
>>@@ -191,18 +263,43 @@ kill_it: {
>>   			dentry_stat.nr_unused--;
>>   		}
>>   		list_del(&dentry->d_child);
>>+		parent = dentry->d_parent;
>>+		dcache_shrinker_add(&lds, parent, dentry);
>> 		dentry_stat.nr_dentry--;	/* For d_free, below */
>> 		/*drops the locks, at that point nobody can reach this dentry */
>> 		dentry_iput(dentry);
>>-		parent = dentry->d_parent;
>> 		d_free(dentry);
>> 		if (dentry == parent)
> 
> 
> Aren't we leaving local variable `lds' on the global list here?
yup, my fault.
You patch solves this.

Kirill