Re: [XEN PATCH] automation/eclair: configure Rule 13.6 and custom service B.UNEVALEFF

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 22/06/2024 11:48 am, Federico Serafini wrote:
> diff --git a/automation/eclair_analysis/ECLAIR/deviations.ecl 
> b/automation/eclair_analysis/ECLAIR/deviations.ecl
> index e2653f77eb..6bdfe7c883 100644
> --- a/automation/eclair_analysis/ECLAIR/deviations.ecl
> +++ b/automation/eclair_analysis/ECLAIR/deviations.ecl
> @@ -328,6 +328,16 @@ of the short-circuit evaluation strategy of such logical 
> operators."
>  -config=MC3R1.R13.5,reports+={disapplied,"any()"}
>  -doc_end
>  
> +-doc_begin="Macros alternative_v?call[0-9] use sizeof and typeof to check 
> that th argument types match the corresponding parameter ones."

Typo "th" => "the".  Can be fixed on commit.

> +-config=MC3R1.R13.6,reports+={deliberate,"any_area(any_loc(any_exp(macro(^alternative_vcall[0-9]$))&&file(^xen/arch/x86.*$)))"}
> +-config=B.UNEVALEFF,reports+={deliberate,"any_area(any_loc(any_exp(macro(^alternative_v?call[0-9]$))&&file(^xen/arch/x86.*$)))"}

There will be expansions of these macros outside of arch/x86/. 
drivers/iommu/ as an example.

Is the file() clause targetting the source of the macro (in which it
probably wants making more specific to alternative_call.h), or the
expansions of the macro (in which case it probably wants dropping
entirely) ?

> +-doc_end
> +
> +-doc_begin="Macro chk_fld is only used to introduce BUILD_BUG_ON checks in 
> very specific cases where the usage is correct and can be checked by code 
> inspection.
> +The BUILD_BUG_ON checks check that EFI_TIME and struct xenpf_efi_time fields 
> match."
> +-config=MC3R1.R13.6,reports+={deliberate,"any_area(any_loc(any_exp(macro(^chk_fld$))&&file(^xen/common/efi/runtime\\.c$)))"}
> +-doc_end

It's just occurred to me.  Anything, no matter how complicated, inside a
BUILD_BUG_ON() is clearly a compile time thing so has no relevant side
effects.

Is it possible to generally exclude any sizeof/alignof/typeof inside a
BUILD_BUG_ON()?  That would be better than identifying every user locally.

~Andrew