Re: [CHECKER] 120 potential dereference to invalid pointers errors forlinux 2.4.1

[Keith Owens <kaos@xxxxxxxxxx>]

On Sun, 18 Mar 2001 06:29:50 -0500, 
Jeff Garzik <jgarzik@xxxxxxxxxxxxxxxx> wrote:
>Junfeng Yang wrote:
>> Start --->
>>             busy = kmalloc(sizeof(erase_busy_t), GFP_KERNEL);
>> Error --->
>
>This sizeof() construct may be a special case for your checker, but it's
>a common one for the kernel...  It definitely doesn't de-reference a
>pointer.

IMHO the above line is a bad construct.  If the type of the variable
changes it is extremely easy to miss the fact that *alloc is now
returning the wrong size.  I always do

        busy = kmalloc(sizeof(*busy), GFP_KERNEL);

and let the compiler insert the correct type.

For the checker, you can also have typeof().  kdb has this line

        typeof (*ef)    local_ef;

The type definition of ef is kdb_eframe_t which is "pointer to some
arch dependent type" and local_ef is in arch independent code, much
easier to do this than use multiple #ifdef.  Of course it would have
been even easier if kdb had separate types for the struct and the
pointer to the struct, then I would not need typeof().  OTOH I am sure
that somebody will find a use for typeof().