]> git.kernelconcepts.de Git - karo-tx-linux.git/blobdiff - kernel/bpf/hashtab.c
tracing: Fix stacktrace skip depth in trace_buffer_unlock_commit_regs()
[karo-tx-linux.git] / kernel / bpf / hashtab.c
index 19909b22b4f8398d7373cc3838b835728265ed23..34777b3746fadf6407ca9805f33df608be619f21 100644 (file)
@@ -64,12 +64,35 @@ static struct bpf_map *htab_map_alloc(union bpf_attr *attr)
                 */
                goto free_htab;
 
-       err = -ENOMEM;
+       if (htab->map.value_size >= (1 << (KMALLOC_SHIFT_MAX - 1)) -
+           MAX_BPF_STACK - sizeof(struct htab_elem))
+               /* if value_size is bigger, the user space won't be able to
+                * access the elements via bpf syscall. This check also makes
+                * sure that the elem_size doesn't overflow and it's
+                * kmalloc-able later in htab_map_update_elem()
+                */
+               goto free_htab;
+
+       htab->elem_size = sizeof(struct htab_elem) +
+                         round_up(htab->map.key_size, 8) +
+                         htab->map.value_size;
+
        /* prevent zero size kmalloc and check for u32 overflow */
        if (htab->n_buckets == 0 ||
            htab->n_buckets > U32_MAX / sizeof(struct hlist_head))
                goto free_htab;
 
+       if ((u64) htab->n_buckets * sizeof(struct hlist_head) +
+           (u64) htab->elem_size * htab->map.max_entries >=
+           U32_MAX - PAGE_SIZE)
+               /* make sure page count doesn't overflow */
+               goto free_htab;
+
+       htab->map.pages = round_up(htab->n_buckets * sizeof(struct hlist_head) +
+                                  htab->elem_size * htab->map.max_entries,
+                                  PAGE_SIZE) >> PAGE_SHIFT;
+
+       err = -ENOMEM;
        htab->buckets = kmalloc_array(htab->n_buckets, sizeof(struct hlist_head),
                                      GFP_USER | __GFP_NOWARN);
 
@@ -85,13 +108,6 @@ static struct bpf_map *htab_map_alloc(union bpf_attr *attr)
        raw_spin_lock_init(&htab->lock);
        htab->count = 0;
 
-       htab->elem_size = sizeof(struct htab_elem) +
-                         round_up(htab->map.key_size, 8) +
-                         htab->map.value_size;
-
-       htab->map.pages = round_up(htab->n_buckets * sizeof(struct hlist_head) +
-                                  htab->elem_size * htab->map.max_entries,
-                                  PAGE_SIZE) >> PAGE_SHIFT;
        return &htab->map;
 
 free_htab:
@@ -222,7 +238,7 @@ static int htab_map_update_elem(struct bpf_map *map, void *key, void *value,
        WARN_ON_ONCE(!rcu_read_lock_held());
 
        /* allocate new element outside of lock */
-       l_new = kmalloc(htab->elem_size, GFP_ATOMIC);
+       l_new = kmalloc(htab->elem_size, GFP_ATOMIC | __GFP_NOWARN);
        if (!l_new)
                return -ENOMEM;