Re: [RFC][PATCH] x86: dovetail: Permit to declare a trap handled

[Richard Weinberger <richard@sigma-star.at>]

Hi!

Sorry for exhuming this old mail thread.

Am Donnerstag, 10. November 2022, 10:00:33 CET schrieb Philippe Gerum:
> >> >>> I also did some similar changes for arm64. As we also use imx8 boards.
> >> >>> If there is interest in those, I could post them.
> >> >>
> >> >> Yes, please share. Then we can continue with proposing it for the latest
> >> >> dovetail port (6.1 now).
> >> >>
> >> >
> >> > There is a slightly different approach which expresses the intent of
> >> > that notification hook Dovetail-wise:
> >> >
> >> > - if the real-time core switches in-band upon notification, then the
> >> >   in-band kernel code may assume that all regular fixups need to be
> >> >   applied next. This is the common case for most traps.
> >> >
> >> > - if the real-time core stays on the out-of-band stage upon
> >> >   notification, then the in-band kernel code should assume that all
> >> >   required fixups were done from that stage already, returning asap from
> >> >   the outer trap handler.
> >> >
> >> > This gives the same flexibility than by expecting the hook to return a
> >> > continuation status, but makes sure that regular fixups cannot be
> >> > bypassed when/if switching to in-band mode is required to handle the
> >> > trap. Rationale: the regular in-band trap handlers know better about the
> >> > way to deal with those events from in-band context.
> >> >
> >
> > I like that idea because as I was looking at the arm code, I was not sure what would happen if we are out of band but 
> > indicate that the kernel should handle the exception. With your idea that’s not possible.
> >
> > Will have a look at my arm64 patch and try to adapt it.
> >
> 
> Ok, so since the idea is generally agreed, I'll push the x86 and arm
> support for this in a couple of days.

Currently, I'm looking into implementing Johannes's RT-Signals for ARM64,
but from what I see on the dovetail side, only x86 has gained support to declare traps handled.

At least on x86, I see that a trap is considered handled if no switch to in-band has occurred.
Did I miss something?

Thanks,
//richard

-- 
​​​​​sigma star gmbh | Eduard-Bodem-Gasse 6, 6020 Innsbruck, AUT
UID/VAT Nr: ATU 66964118 | FN: 374287y