bpf: Register dtor for freeing special fields

There is a race window where BPF hash map elements can leak special
fields if the program with access to the map value recreates these
special fields between the check_and_free_fields done on the map value
and its eventual return to the memory allocator.

Several ways were explored prior to this patch, most notably [0] tried
to use a poison value to reject attempts to recreate special fields for
map values that have been logically deleted but still accessible to BPF
programs (either while sitting in the free list or when reused). While
this approach works well for task work, timers, wq, etc., it is harder
to apply the idea to kptrs, which have a similar race and failure mode.

Instead, we change bpf_mem_alloc to allow registering destructor for
allocated elements, such that when they are returned to the allocator,
any special fields created while they were accessible to programs in the
mean time will be freed. If these values get reused, we do not free the
fields again before handing the element back. The special fields thus
may remain initialized while the map value sits in a free list.

When bpf_mem_alloc is retired in the future, a similar concept can be
introduced to kmalloc_nolock-backed kmem_cache, paired with the existing
idea of a constructor.

Note that the destructor registration happens in map_check_btf, after
the BTF record is populated and (at that point) avaiable for inspection
and duplication. Duplication is necessary since the freeing of embedded
bpf_mem_alloc can be decoupled from actual map lifetime due to logic
introduced to reduce the cost of rcu_barrier()s in mem alloc free path in
9f2c6e96c65e ("bpf: Optimize rcu_barrier usage between hash map and bpf_mem_alloc.").

As such, once all callbacks are done, we must also free the duplicated
record. To remove dependency on the bpf_map itself, also stash the key
size of the map to obtain value from htab_elem long after the map is
gone.

  [0]: https://lore.kernel.org/bpf/20260216131341.1285427-1-mykyta.yatsenko5@gmail.com

Fixes: 14a324f6a67e ("bpf: Wire up freeing of referenced kptr")
Fixes: 1bfbc267ec91 ("bpf: Enable bpf_timer and bpf_wq in any context")
Reported-by: Alexei Starovoitov <ast@kernel.org>
Tested-by: syzbot@syzkaller.appspotmail.com
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20260227224806.646888-2-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

2 files changed, 134 insertions, 11 deletions

diff --git a/kernel/bpf/hashtab.c b/kernel/bpf/hashtab.c
index 3b9d297a53be..582f0192b7e1 100644
--- a/kernel/bpf/hashtab.c
+++ b/kernel/bpf/hashtab.c
@@ -125,6 +125,11 @@ struct htab_elem {
 	char key[] __aligned(8);
 };
 
+struct htab_btf_record {
+	struct btf_record *record;
+	u32 key_size;
+};
+
 static inline bool htab_is_prealloc(const struct bpf_htab *htab)
 {
 	return !(htab->map.map_flags & BPF_F_NO_PREALLOC);
@@ -457,6 +462,84 @@ static int htab_map_alloc_check(union bpf_attr *attr)
 	return 0;
 }
 
+static void htab_mem_dtor(void *obj, void *ctx)
+{
+	struct htab_btf_record *hrec = ctx;
+	struct htab_elem *elem = obj;
+	void *map_value;
+
+	if (IS_ERR_OR_NULL(hrec->record))
+		return;
+
+	map_value = htab_elem_value(elem, hrec->key_size);
+	bpf_obj_free_fields(hrec->record, map_value);
+}
+
+static void htab_pcpu_mem_dtor(void *obj, void *ctx)
+{
+	void __percpu *pptr = *(void __percpu **)obj;
+	struct htab_btf_record *hrec = ctx;
+	int cpu;
+
+	if (IS_ERR_OR_NULL(hrec->record))
+		return;
+
+	for_each_possible_cpu(cpu)
+		bpf_obj_free_fields(hrec->record, per_cpu_ptr(pptr, cpu));
+}
+
+static void htab_dtor_ctx_free(void *ctx)
+{
+	struct htab_btf_record *hrec = ctx;
+
+	btf_record_free(hrec->record);
+	kfree(ctx);
+}
+
+static int htab_set_dtor(const struct bpf_htab *htab, void (*dtor)(void *, void *))
+{
+	u32 key_size = htab->map.key_size;
+	const struct bpf_mem_alloc *ma;
+	struct htab_btf_record *hrec;
+	int err;
+
+	/* No need for dtors. */
+	if (IS_ERR_OR_NULL(htab->map.record))
+		return 0;
+
+	hrec = kzalloc(sizeof(*hrec), GFP_KERNEL);
+	if (!hrec)
+		return -ENOMEM;
+	hrec->key_size = key_size;
+	hrec->record = btf_record_dup(htab->map.record);
+	if (IS_ERR(hrec->record)) {
+		err = PTR_ERR(hrec->record);
+		kfree(hrec);
+		return err;
+	}
+	ma = htab_is_percpu(htab) ? &htab->pcpu_ma : &htab->ma;
+	/* Kinda sad, but cast away const-ness since we change ma->dtor. */
+	bpf_mem_alloc_set_dtor((struct bpf_mem_alloc *)ma, dtor, htab_dtor_ctx_free, hrec);
+	return 0;
+}
+
+static int htab_map_check_btf(const struct bpf_map *map, const struct btf *btf,
+			      const struct btf_type *key_type, const struct btf_type *value_type)
+{
+	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
+
+	if (htab_is_prealloc(htab))
+		return 0;
+	/*
+	 * We must set the dtor using this callback, as map's BTF record is not
+	 * populated in htab_map_alloc(), so it will always appear as NULL.
+	 */
+	if (htab_is_percpu(htab))
+		return htab_set_dtor(htab, htab_pcpu_mem_dtor);
+	else
+		return htab_set_dtor(htab, htab_mem_dtor);
+}
+
 static struct bpf_map *htab_map_alloc(union bpf_attr *attr)
 {
 	bool percpu = (attr->map_type == BPF_MAP_TYPE_PERCPU_HASH ||
@@ -2281,6 +2364,7 @@ const struct bpf_map_ops htab_map_ops = {
 	.map_seq_show_elem = htab_map_seq_show_elem,
 	.map_set_for_each_callback_args = map_set_for_each_callback_args,
 	.map_for_each_callback = bpf_for_each_hash_elem,
+	.map_check_btf = htab_map_check_btf,
 	.map_mem_usage = htab_map_mem_usage,
 	BATCH_OPS(htab),
 	.map_btf_id = &htab_map_btf_ids[0],
@@ -2303,6 +2387,7 @@ const struct bpf_map_ops htab_lru_map_ops = {
 	.map_seq_show_elem = htab_map_seq_show_elem,
 	.map_set_for_each_callback_args = map_set_for_each_callback_args,
 	.map_for_each_callback = bpf_for_each_hash_elem,
+	.map_check_btf = htab_map_check_btf,
 	.map_mem_usage = htab_map_mem_usage,
 	BATCH_OPS(htab_lru),
 	.map_btf_id = &htab_map_btf_ids[0],
@@ -2482,6 +2567,7 @@ const struct bpf_map_ops htab_percpu_map_ops = {
 	.map_seq_show_elem = htab_percpu_map_seq_show_elem,
 	.map_set_for_each_callback_args = map_set_for_each_callback_args,
 	.map_for_each_callback = bpf_for_each_hash_elem,
+	.map_check_btf = htab_map_check_btf,
 	.map_mem_usage = htab_map_mem_usage,
 	BATCH_OPS(htab_percpu),
 	.map_btf_id = &htab_map_btf_ids[0],
@@ -2502,6 +2588,7 @@ const struct bpf_map_ops htab_lru_percpu_map_ops = {
 	.map_seq_show_elem = htab_percpu_map_seq_show_elem,
 	.map_set_for_each_callback_args = map_set_for_each_callback_args,
 	.map_for_each_callback = bpf_for_each_hash_elem,
+	.map_check_btf = htab_map_check_btf,
 	.map_mem_usage = htab_map_mem_usage,
 	BATCH_OPS(htab_lru_percpu),
 	.map_btf_id = &htab_map_btf_ids[0],
diff --git a/kernel/bpf/memalloc.c b/kernel/bpf/memalloc.c
index bd45dda9dc35..682a9f34214b 100644
--- a/kernel/bpf/memalloc.c
+++ b/kernel/bpf/memalloc.c
@@ -102,6 +102,8 @@ struct bpf_mem_cache {
 	int percpu_size;
 	bool draining;
 	struct bpf_mem_cache *tgt;
+	void (*dtor)(void *obj, void *ctx);
+	void *dtor_ctx;
 
 	/* list of objects to be freed after RCU GP */
 	struct llist_head free_by_rcu;
@@ -260,12 +262,14 @@ static void free_one(void *obj, bool percpu)
 	kfree(obj);
 }
 
-static int free_all(struct llist_node *llnode, bool percpu)
+static int free_all(struct bpf_mem_cache *c, struct llist_node *llnode, bool percpu)
 {
 	struct llist_node *pos, *t;
 	int cnt = 0;
 
 	llist_for_each_safe(pos, t, llnode) {
+		if (c->dtor)
+			c->dtor((void *)pos + LLIST_NODE_SZ, c->dtor_ctx);
 		free_one(pos, percpu);
 		cnt++;
 	}
@@ -276,7 +280,7 @@ static void __free_rcu(struct rcu_head *head)
 {
 	struct bpf_mem_cache *c = container_of(head, struct bpf_mem_cache, rcu_ttrace);
 
-	free_all(llist_del_all(&c->waiting_for_gp_ttrace), !!c->percpu_size);
+	free_all(c, llist_del_all(&c->waiting_for_gp_ttrace), !!c->percpu_size);
 	atomic_set(&c->call_rcu_ttrace_in_progress, 0);
 }
 
@@ -308,7 +312,7 @@ static void do_call_rcu_ttrace(struct bpf_mem_cache *c)
 	if (atomic_xchg(&c->call_rcu_ttrace_in_progress, 1)) {
 		if (unlikely(READ_ONCE(c->draining))) {
 			llnode = llist_del_all(&c->free_by_rcu_ttrace);
-			free_all(llnode, !!c->percpu_size);
+			free_all(c, llnode, !!c->percpu_size);
 		}
 		return;
 	}
@@ -417,7 +421,7 @@ static void check_free_by_rcu(struct bpf_mem_cache *c)
 	dec_active(c, &flags);
 
 	if (unlikely(READ_ONCE(c->draining))) {
-		free_all(llist_del_all(&c->waiting_for_gp), !!c->percpu_size);
+		free_all(c, llist_del_all(&c->waiting_for_gp), !!c->percpu_size);
 		atomic_set(&c->call_rcu_in_progress, 0);
 	} else {
 		call_rcu_hurry(&c->rcu, __free_by_rcu);
@@ -635,13 +639,13 @@ static void drain_mem_cache(struct bpf_mem_cache *c)
 	 * Except for waiting_for_gp_ttrace list, there are no concurrent operations
 	 * on these lists, so it is safe to use __llist_del_all().
 	 */
-	free_all(llist_del_all(&c->free_by_rcu_ttrace), percpu);
-	free_all(llist_del_all(&c->waiting_for_gp_ttrace), percpu);
-	free_all(__llist_del_all(&c->free_llist), percpu);
-	free_all(__llist_del_all(&c->free_llist_extra), percpu);
-	free_all(__llist_del_all(&c->free_by_rcu), percpu);
-	free_all(__llist_del_all(&c->free_llist_extra_rcu), percpu);
-	free_all(llist_del_all(&c->waiting_for_gp), percpu);
+	free_all(c, llist_del_all(&c->free_by_rcu_ttrace), percpu);
+	free_all(c, llist_del_all(&c->waiting_for_gp_ttrace), percpu);
+	free_all(c, __llist_del_all(&c->free_llist), percpu);
+	free_all(c, __llist_del_all(&c->free_llist_extra), percpu);
+	free_all(c, __llist_del_all(&c->free_by_rcu), percpu);
+	free_all(c, __llist_del_all(&c->free_llist_extra_rcu), percpu);
+	free_all(c, llist_del_all(&c->waiting_for_gp), percpu);
 }
 
 static void check_mem_cache(struct bpf_mem_cache *c)
@@ -680,6 +684,9 @@ static void check_leaked_objs(struct bpf_mem_alloc *ma)
 
 static void free_mem_alloc_no_barrier(struct bpf_mem_alloc *ma)
 {
+	/* We can free dtor ctx only once all callbacks are done using it. */
+	if (ma->dtor_ctx_free)
+		ma->dtor_ctx_free(ma->dtor_ctx);
 	check_leaked_objs(ma);
 	free_percpu(ma->cache);
 	free_percpu(ma->caches);
@@ -1014,3 +1021,32 @@ int bpf_mem_alloc_check_size(bool percpu, size_t size)
 
 	return 0;
 }
+
+void bpf_mem_alloc_set_dtor(struct bpf_mem_alloc *ma, void (*dtor)(void *obj, void *ctx),
+			    void (*dtor_ctx_free)(void *ctx), void *ctx)
+{
+	struct bpf_mem_caches *cc;
+	struct bpf_mem_cache *c;
+	int cpu, i;
+
+	ma->dtor_ctx_free = dtor_ctx_free;
+	ma->dtor_ctx = ctx;
+
+	if (ma->cache) {
+		for_each_possible_cpu(cpu) {
+			c = per_cpu_ptr(ma->cache, cpu);
+			c->dtor = dtor;
+			c->dtor_ctx = ctx;
+		}
+	}
+	if (ma->caches) {
+		for_each_possible_cpu(cpu) {
+			cc = per_cpu_ptr(ma->caches, cpu);
+			for (i = 0; i < NUM_CACHES; i++) {
+				c = &cc->cache[i];
+				c->dtor = dtor;
+				c->dtor_ctx = ctx;
+			}
+		}
+	}
+}