[SystemSafety] Comparing reliability predictions with reality

[Paul Sherwood]

On 2025-02-28 07:39, Nick Tudor wrote:
> I understand from studies undertaken with people on this list, that
> some companies make available firmware for their processors. They may
> make little or no claims for the correctness of that very low level
> code and in assessing the repositories for some of this it has been
> found to have little evidence to support making claims for
> correctness. The disclaimer (if there is one), is that this is
> supplied as examples and it is for the user to build their own or/and
> verify themselves. However, on checking, quite often these are used
> ‘out of the box’.

For modern microprocessors (as opposed to microcontrollers), my 
understanding is that **all** vendors supply firmware, but usually only 
as binary blobs for the customers. Typically there is no possibility of 
modification by the users, nor opportunity to verify what the firmware 
is supposed to do - in fact much of what it does may be labelled 
vendor-secret. Note this is not a small amount of code - we guesstimate 
it could typically be ~ 1MLOC.

The disclaimer/examples approach is often applied for the "BSP" e.g. 
boot loader, kernel, drivers etc. which are supplied in addition.

br
Paul