From:   George Prekas <george@enfabrica.net>
To:     linux-kernel@vger.kernel.org, linux-mm@kvack.org
Cc:     Catalin Marinas <catalin.marinas@arm.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Christoph Lameter <cl@linux.com>,
        Pekka Enberg <penberg@kernel.org>,
        David Rientjes <rientjes@google.com>,
        Joonsoo Kim <iamjoonsoo.kim@lge.com>,
        Vlastimil Babka <vbabka@suse.cz>,
        Roman Gushchin <roman.gushchin@linux.dev>,
        Hyeonggon Yoo <42.hyeyoo@gmail.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        "Eric W. Biederman" <ebiederm@xmission.com>,
        Sebastian Andrzej Siewior <bigeasy@linutronix.de>,
        Andy Lutomirski <luto@kernel.org>,
        Peter Zijlstra <peterz@infradead.org>,
        "Liam R. Howlett" <Liam.Howlett@Oracle.com>,
        Fenghua Yu <fenghua.yu@intel.com>,
        Andrei Vagin <avagin@gmail.com>,
        George Prekas <george@enfabrica.net>
Subject: [PATCH 6/9] mm: kmemleak: do not scan cpu_cache of struct kmem_cache
Date:   Mon, 23 Jan 2023 11:04:16 -0600
Message-Id: <20230123170419.7292-7-george@enfabrica.net>
In-Reply-To: <20230123170419.7292-1-george@enfabrica.net>
References: <20230123170419.7292-1-george@enfabrica.net>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

The code already makes sure that kmemleak will not scan similar caches:
array_cache and alien_cache. For the cpu_cache, the code takes a
different approach using kmemleak_erase. This approach handles object
allocations but does not handle transfers to other caches and leads to
undetected leaks.

According to the comment in alloc_arraycache: [...] when such objects
are allocated or transferred to another cache the pointers are not
cleared and they could be counted as valid references during a kmemleak
scan. Therefore, kmemleak must not scan such objects.

Signed-off-by: George Prekas <george@enfabrica.net>
---
 mm/slab.c | 13 +------------
 1 file changed, 1 insertion(+), 12 deletions(-)

diff --git a/mm/slab.c b/mm/slab.c
index 29300fc1289a..a927e1a285d1 100644
--- a/mm/slab.c
+++ b/mm/slab.c
@@ -1744,6 +1744,7 @@ static struct array_cache __percpu *alloc_kmem_cache_cpus(
 	for_each_possible_cpu(cpu) {
 		init_arraycache(per_cpu_ptr(cpu_cache, cpu),
 				entries, batchcount);
+		kmemleak_no_scan(per_cpu_ptr(cpu_cache, cpu));
 	}
 
 	return cpu_cache;
@@ -3023,20 +3024,8 @@ static inline void *____cache_alloc(struct kmem_cache *cachep, gfp_t flags)
 
 	STATS_INC_ALLOCMISS(cachep);
 	objp = cache_alloc_refill(cachep, flags);
-	/*
-	 * the 'ac' may be updated by cache_alloc_refill(),
-	 * and kmemleak_erase() requires its correct value.
-	 */
-	ac = cpu_cache_get(cachep);
 
 out:
-	/*
-	 * To avoid a false negative, if an object that is in one of the
-	 * per-CPU caches is leaked, we need to make sure kmemleak doesn't
-	 * treat the array pointers as a reference to the object.
-	 */
-	if (objp)
-		kmemleak_erase(&ac->entry[ac->avail]);
 	return objp;
 }
 
-- 
2.37.1