Re: [CHECKER] potential dereference of user pointer errors

[Raja R Harinath <harinath@cs.umn.edu>]

Hi,

"David S. Miller" <davem@redhat.com> writes:

> On Mon, 2003-03-24 at 14:28, Raja R Harinath wrote:
>> Something like
>> 
>>   /* struct user_space should never be defined.  */
>>   typedef struct user_space user_space;
>> 
>>   int copy_to_user (user_space *to, char *from, size_t len);
>>   int copy_from_user (char *to, user_space *from, size_t len);
>>   /* ... */
>> 
>>   #define TREAT_AS_USER_SPACE_POINTER(p) \
>>             ({                                  \
>>               BUG_ON(get_fs() != get_gs());     \
>>               (user_space *)p;                  \
>>             })
>
> A great idea, we'd need to use this struct user_space at every
> system call, 

Which is good, I think.  It annotates the source.

> and it doesn't work to well when pointers are embedded inside of a
> structure.

I don't understand this point.  'struct user_space' will never be
defined.  The C compiler ensure that 'struct user_space *p' cannot be
dereferenced then, since it will be a pointer to an incomplete type.

I don't see any difference between

  struct foo { char *p; };

and 

  struct foo { user_space *p; };

in terms of the code generated.  (On second reading, I don't think
this was your point though).

- Hari
-- 
Raja R Harinath ------------------------------ harinath@cs.umn.edu