orangefs: silence harmless integer overflow warning
authorDan Carpenter <dan.carpenter@oracle.com>
Sat, 21 Jan 2017 05:04:45 +0000 (08:04 +0300)
committerMike Marshall <hubcap@omnibond.com>
Fri, 3 Feb 2017 19:37:15 +0000 (14:37 -0500)
The issue here is that in orangefs_bufmap_alloc() we do:

bufmap->buffer_index_array =
kzalloc(DIV_ROUND_UP(bufmap->desc_count, BITS_PER_LONG), GFP_KERNEL);

If we choose a bufmap->desc_count like -31 then it means the
DIV_ROUND_UP ends up having an integer overflow.   The result is that
kzalloc() returns the ZERO_SIZE_PTR and there is a static checker
warning.

But this bug is harmless because on the next lines we use ->desc_count
to do a kcalloc().  That has integer overflow checking built in so the
kcalloc() fails and we return an error code.

Anyway, it doesn't make sense to talk about negative sizes and blocking
them silences the static checker warning.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Mike Marshall <hubcap@omnibond.com>
fs/orangefs/orangefs-bufmap.c

index 75375e90a63f32e4bac0ef3e996c960c673b250e..6333cbbdfef7ae652c1a4e6c4d2818ae1cd188d7 100644 (file)
@@ -344,6 +344,11 @@ int orangefs_bufmap_initialize(struct ORANGEFS_dev_map_desc *user_desc)
                     user_desc->size,
                     user_desc->count);
 
+       if (user_desc->total_size < 0 ||
+           user_desc->size < 0 ||
+           user_desc->count < 0)
+               goto out;
+
        /*
         * sanity check alignment and size of buffer that caller wants to
         * work with