Re: [RESEND PATCH v2] params: Annotate struct module_param_attrs with __counted_by()

[Andy Shevchenko]

On Fri, Sep 13, 2024 at 10:09:25PM +0300, Andy Shevchenko wrote:
> On Fri, Sep 13, 2024 at 09:03:06PM +0200, Thorsten Blum wrote:
> > On 13. Sep 2024, at 20:40, Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx> wrote:
> > > On Fri, Sep 13, 2024 at 09:46:30AM -0700, Nathan Chancellor wrote:
> > >> On Mon, Sep 09, 2024 at 06:27:26PM +0200, Thorsten Blum wrote:
> > >>> Add the __counted_by compiler attribute to the flexible array member
> > >>> attrs to improve access bounds-checking via CONFIG_UBSAN_BOUNDS and
> > >>> CONFIG_FORTIFY_SOURCE.
> > >>> 
> > >>> Increment num before adding a new param_attribute to the attrs array and
> > >>> adjust the array index accordingly. Increment num immediately after the
> > >>> first reallocation such that the reallocation for the NULL terminator
> > >>> only needs to add 1 (instead of 2) to mk->mp->num.
> > >>> 
> > >>> Use struct_size() instead of manually calculating the size for the
> > >>> reallocation.
> > >>> 
> > >>> Use krealloc_array() for the additional NULL terminator.
> > > 
> > >>> /* Fix up all the pointers, since krealloc can move us */
> > >>> for (i = 0; i < mk->mp->num; i++)
> > > 
> > > Shouldn't this for loop and followed by assignment also be -1:ed?
> > 
> > That should be fine as mk->mp->num was already incremented before the
> > for-loop.
> 
> Exactly my point. This is behavioural change AFAICS as the original code used
> the old (-1:ed) value, no?

Ah, I see now. It was in the original code, but closer to that loop...


-- 
With Best Regards,
Andy Shevchenko