On 04.08.2022 21:46, Andrew Cooper wrote:
> On 04/08/2022 16:04, Jane Malalane wrote:
>> RFC: I'm unsure on where the page_fault symbol should end, i.e. if
>> unlike current code handle_exception_saved should be within page_fault
>> like handle_exception is or not.
>
> Jan: we've got two examples (page fault, and NMI) which don't form any
> reasonable function layout. Both of these are fallthrough into
> handle_{ist,}_exception.
>
> I suggested labelling handle_{ist,}_exception as the main symbol, and
> keeping {page_fault,nmi} as small stubs, because we want backtraces to
> stay the same and not report {page_fault,nmi} for everything.
I.e. the opposite of what the patch currently does. That's fine with me
in principle (sadly there's no STT_THUNK or alike, which might allow
better reflecting the purpose yet still not marking these as STT_FUNC
nor leaving them at STT_NOTYPE), but the small stubs then want an end
annotation, so that their code is covered by some [start,start+size)
pair in the symbol table. IOW I think that as a final result (not
necessarily right after this series) we want all code and data
contributions to be covered by such a range. Which in turn means for
this series that _if_ an area is touched, it should be brought into
that intended shape.
Jan
