Re: [Xen-devel] [RFC 5/6] arm64: call enter_hypervisor_head only when it is needed

[Andrii Anisov <andrii.anisov@xxxxxxxxx>]

On 26.07.19 13:59, Julien Grall wrote:
> Hi,
>
> On 26/07/2019 11:37, Andrii Anisov wrote:
> > From: Andrii Anisov <andrii_anisov@xxxxxxxx>
> >
> > On ARM64 we know exactly if trap happened from hypervisor or guest, so
> > we do not need to take that decision. This reduces a condition for
> > all enter_hypervisor_head calls and the function call for traps from
> > the hypervisor mode.
> One condition lost but ...
...In the hot path (actually at any trap).
And the function call for traps from hyp.

> > Currently, it is implemented for ARM64 only. Integrating the stuff
> > with ARM32 requires moving ` if ( guest_mode(regs) )` condition
> > into ARM32 specific traps.
> ... one more divergence between arm32 and arm64.
>
> There are probably dozens of more conditions in the code that are not necessary 
> for one of the architectures.
> Yet there are value to keep everything common because the benefits outweigh the 
> likely non performance improvement.
I'm not seeing any benefits in calling `enter_hypervisor_head()` from functions 
named `do_trap_hyp_*`. That code is confusing for me.
IMHO, dividing `do_trap_irq/fiq()` into guest and hyp specific functions is not 
a big deal. Even for ARM32. Moreover, it will make more obvious the fact that 
nothing from `enter_hypervisor_head()` is done for IRQ traps taken from hyp.

> So I am not convinced that this has any value for Xen.
OK, I wouldn't insist.

--
Sincerely,
Andrii Anisov.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel