Re: [patch 13/16] x86/entry: Move irqflags and context tracking to C for simple idtentries

[Thomas Gleixner]

Andy Lutomirski <luto@xxxxxxxxxxxxxx> writes:
>> On Feb 26, 2020, at 8:28 AM, Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
>> 
>> ïOn Wed, Feb 26, 2020 at 07:11:39AM -0800, Andy Lutomirski wrote:
>> 
>>> In some sense, this is a weakness of the magic macro approach.  Some
>>> of the entries just want to have code that runs before all the entry
>>> fixups.  This is an example of it.  So are the cr2 reads.  It can all
>>> be made to work, but it's a bit gross.
>> 
>> Right. In my current pile (new patche since last posting) I also have
>> one that makes #DB save-clear/restore DR7.
>> 
>> I got it early enough that only a watchpoint on the task stack can still
>> screw us over, since I also included your patch that excludes
>> cpu_entry_area.
>
> Hmm. It would be nice to prevent watchpoints on the task stack, but that would need some trickery.  It could be done.
>
>> 
>> Pushing it earlier still would require calling into C from the entry
>> stack, which I know is on your todo list, but we're not quite there yet.
>
> Indeed.
>
> This is my main objection to the DEFINE_IDTENTRY stuff. Itâs *great*
> for the easy cases, but itâs not so great for the nasty cases. Maybe
> we should open code PF, MC, DB, etc.  (And kill the kvm special case
> for PF.  I have a working patch for that and I can send it.)

I'm fine with that. I like the obvious easy ones as it spares to
duplicate all the same crap all over the place.

Making the nasty ones have:

#define DEFINE_IDTENTRY_$NASTY(exc_name)	\
__visible void notrace exc_name(args....)

would solve this and you can stick whatever you want into it and have
the NOPROBE added manually. Hmm?

Thanks,

        tglx