Re: [Minios-devel] [UNIKRAFT RFC PATCH 2/2] plat/common: Add a trap function to handle Arm64 irq trap

[Julien Grall <julien.grall@xxxxxxx>]

On 11/10/18 2:02 PM, Jianyong Wu (Arm Technology China) wrote:
> Hi,
Hi,

> > -----Original Message-----
> > From: Julien Grall <julien.grall@xxxxxxx>
> > Sent: Saturday, November 10, 2018 1:56 AM
> > To: Jianyong Wu (Arm Technology China) <Jianyong.Wu@xxxxxxx>; minios-
> > devel@xxxxxxxxxxxxxxxxxxxx; simon.kuenzer@xxxxxxxxx
> > Cc: Kaly Xin (Arm Technology China) <Kaly.Xin@xxxxxxx>; nd
> > <nd@xxxxxxx>; Wei Chen (Arm Technology China) <Wei.Chen@xxxxxxx>
> > Subject: Re: [Minios-devel] [UNIKRAFT RFC PATCH 2/2] plat/common: Add a
> > trap function to handle Arm64 irq trap
> >
> > Hi,
> >
> > On 09/11/2018 09:02, Jianyong Wu wrote:
> > > From: Wei Chen <wei.chen@xxxxxxx>
> > >
> > > Change-Id: I3652599f8385fc67a29ab2d9e67861590c0f9f6c
> > > Signed-off-by: Wei Chen <wei.chen@xxxxxxx>
> > > ---
> > >    plat/common/arm/traps.c   | 27 +++++++++++++++++++++++++++
> > >    plat/kvm/arm/exceptions.S | 10 +++++++++-
> > >    2 files changed, 36 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/plat/common/arm/traps.c b/plat/common/arm/traps.c index
> > > 43235be..257106b 100644
> > > --- a/plat/common/arm/traps.c
> > > +++ b/plat/common/arm/traps.c
> > > @@ -23,6 +23,7 @@
> > >    #include <string.h>
> > >    #include <uk/print.h>
> > >    #include <uk/assert.h>
> > > +#include <arm/gic-v2.h>
> > >
> > >    static const char *exception_modes[]= {
> > >         "Synchronous Abort",
> > > @@ -68,3 +69,29 @@ void trap_el1_sync(struct __regs *regs, uint64_t far)
> > >         dump_registers(regs, far);
> > >         ukplat_crash();
> > >    }
> > > +
> > > +void trap_el1_irq(struct __regs *regs, uint64_t far) {
> > > +       uint32_t stat, irq;
> > > +
> > > +       do {
> > > +               stat = gic_ack_irq();
> > > +               irq = stat & GICC_IAR_INTID_MASK;
> > > +
> > > +               uk_printd(DLVL_CRIT, "Unikraft: EL1 IRQ#%d trap caught\n",
> > irq);
> > > +
> > > +               /*
> > > +                * TODO: Hanle IPI&SGI interrupts here
> > > +                */
> > > +               if (irq < GIC_MAX_IRQ) {
> > > +                       gic_eoi_irq(stat);
> > > +                       isb();
> > > +                       _ukplat_irq_handle((unsigned long)irq);
> > > +                       continue;
> > > +               }
> > Looking again at this code, could you explain why you EOI the interrupt
> > before handling it?
> I just copy this from freebsd, but I don't know why. Do you know why bsd does 
> that?
I guess you mean you copied from sys/arm/arm/gic.c? I took a look at it 
and also at sys/arm64/arm64/gic_v3.c.
FreeBSD is EOIing the interrupt early (i.e before the handler is called) 
only for edge-interrupt and IPI. For level interrupt, they will be 
acknowledge later. This is because as soon as you EOI a level interrupt, 
it can possibly fire again because the interrupt line is still high/low.
For simplicity, I would always EOI the interrupt after 
_ukplat_irq_handle(...) is called.
Cheers,

--
Julien Grall

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