Re: [patch 2/2] fbcmap: integer overflow bug

[walter harms <wharms@bfs.de>]

Am 15.11.2010 08:20, schrieb Dan Carpenter:
> On Mon, Nov 15, 2010 at 07:56:05AM +0100, Geert Uytterhoeven wrote:
>> On Mon, Nov 15, 2010 at 05:48, Paul Mundt <lethal@linux-sh.org> wrote:
>>> On Sat, Nov 13, 2010 at 01:07:18PM +0300, Dan Carpenter wrote:
>>>> @@ -256,8 +264,12 @@ int fb_set_user_cmap(struct fb_cmap_user *cmap, struct fb_info *info)
>>>>       int rc, size = cmap->len * sizeof(u16);
>>>>       struct fb_cmap umap;
>>>>
>>>> +     if (cmap->len * 2 > INT_MAX)
>>
>> Isn't that another integer overflow? I.e. should be "if (cmap->len >
>> INT_MAX / sizeof(u16))" instead?
>>
> 
> Yeah it is.  :/
> 
> I'll change it to:
> 	if (size < 0 || size < cmap->len)
> like Paul asked.
> 

I do not see the rest of the code but it looks like
cmap->len is size in int8_t. So the upper limit is something like
INT_MAX/(sizeof(u16)*2). Perhaps we can call it a char ?
is there ANY system that has a more than 256 colors in R|G|B ?

<spying for struct fb_cmap_user  >
 __u32 len;
 __u16 __user *red;

So no need to check for <0.

hope that helps,
re,
 wh