Re: [PATCH] LoongArch: Don't disable EIOINTC master core

[Huacai Chen]

Hi, Marc,

On Tue, Aug 9, 2022 at 8:53 PM Marc Zyngier <maz@xxxxxxxxxx> wrote:
>
> On Tue, 09 Aug 2022 11:39:15 +0100,
> Huacai Chen <chenhuacai@xxxxxxxxxx> wrote:
> >
> > Hi, Marc,
> >
> > On Tue, Aug 9, 2022 at 6:20 PM Marc Zyngier <maz@xxxxxxxxxx> wrote:
> > >
> > > On Tue, 09 Aug 2022 10:19:31 +0100,
> > > Huacai Chen <chenhuacai@xxxxxxxxxx> wrote:
> > > >
> > > > Hi, Marc,
> > > >
> > > > On Tue, Aug 9, 2022 at 4:56 PM Marc Zyngier <maz@xxxxxxxxxx> wrote:
> > > > >
> > > > > On Tue, 09 Aug 2022 08:45:22 +0100,
> > > > > Huacai Chen <chenhuacai@xxxxxxxxxxx> wrote:
> > > > > >
> > > > > > This patch fix a CPU hotplug issue. The EIOINTC master core (the first
> > > > > > core of an EIOINTC node) should not be disabled at runtime, since it has
> > > > > > the responsibility of dispatching I/O interrupts.
> > > > > >
> > > > > > Signed-off-by: Huacai Chen <chenhuacai@xxxxxxxxxxx>
> > > > > > ---
> > > > > >  arch/loongarch/kernel/smp.c            | 9 +++++++++
> > > > > >  drivers/irqchip/irq-loongson-eiointc.c | 5 +++++
> > > > > >  2 files changed, 14 insertions(+)
> > > > > >
> > > > > > diff --git a/arch/loongarch/kernel/smp.c b/arch/loongarch/kernel/smp.c
> > > > > > index 09743103d9b3..54901716f8de 100644
> > > > > > --- a/arch/loongarch/kernel/smp.c
> > > > > > +++ b/arch/loongarch/kernel/smp.c
> > > > > > @@ -242,9 +242,18 @@ void loongson3_smp_finish(void)
> > > > > >
> > > > > >  static bool io_master(int cpu)
> > > > > >  {
> > > > > > +     int i, node, master;
> > > > > > +
> > > > > >       if (cpu == 0)
> > > > > >               return true;
> > > > > >
> > > > > > +     for (i = 1; i < loongson_sysconf.nr_io_pics; i++) {
> > > > > > +             node = eiointc_get_node(i);
> > > > > > +             master = cpu_number_map(node * CORES_PER_EIO_NODE);
> > > > > > +             if (cpu == master)
> > > > > > +                     return true;
> > > > > > +     }
> > > > > > +
> > > > > >       return false;
> > > > > >  }
> > > > > >
> > > > > > diff --git a/drivers/irqchip/irq-loongson-eiointc.c b/drivers/irqchip/irq-loongson-eiointc.c
> > > > > > index 170dbc96c7d3..6c99a2ff95f5 100644
> > > > > > --- a/drivers/irqchip/irq-loongson-eiointc.c
> > > > > > +++ b/drivers/irqchip/irq-loongson-eiointc.c
> > > > > > @@ -56,6 +56,11 @@ static void eiointc_enable(void)
> > > > > >       iocsr_write64(misc, LOONGARCH_IOCSR_MISC_FUNC);
> > > > > >  }
> > > > > >
> > > > > > +int eiointc_get_node(int id)
> > > > > > +{
> > > > > > +     return eiointc_priv[id]->node;
> > > > > > +}
> > > > > > +
> > > > > >  static int cpu_to_eio_node(int cpu)
> > > > > >  {
> > > > > >       return cpu_logical_map(cpu) / CORES_PER_EIO_NODE;
> > > > >
> > > > >
> > > > > I don't understand why it has to be this complex and make any use of
> > > > > the node number.
> > > > >
> > > > > As I understand it, CPU-0 in any EIOINTC block is a master. So all you
> > > > > need to find out is whether the CPU number is a multiple of
> > > > > CORES_PER_EIO_NODE.
> > > > CPU-0 in any EIOINTC block may be a master, but not absolutely be a
> > > > master to dispatch I/O interrupts. If there is no bridge under a
> > > > EIOINTC, then this EIOINTC doesn't handle I/O interrupts, and it can
> > > > be disabled at runtime.
> > >
> > > But that's not what your code is checking, is it? You're only
> > > reporting the node number, irrespective of whether there is anything
> > > behind the EIOINTC.
> > The return value of eiointc_get_node() means "this eio-node has a
> > downstream bridge, so the master core of this eio-node cannot be
> > disabled". :)
>
> So what is exactly the meaning of this node? All the EIOINTCs do have
> one (it is coming from ACPI, and taken at face value), so the node
> really is only a proxy for the CPU numbers that are attached to it,
> isn't it? Can you have cores without an EIOINTC?
>
> Now, if this is relevant to the arch code, I'd rather you keep track
> of this directly in the arch code, because it looks really odd to peek
> at an irqchip data structure for something that the core code should
> have the first place.
Emm, yes, you are right, this problem seems can be solved by only
touching the arch code. Thanks.

Huacai
>
> It also strikes me that this patch has *zero* effect, as nothing ever
> sets loongson_sysconf.nr_io_pics. Try this:
>
> diff --git a/arch/loongarch/include/asm/bootinfo.h b/arch/loongarch/include/asm/bootinfo.h
> index 9b8d49d9e61b..13e5e5e21ffd 100644
> --- a/arch/loongarch/include/asm/bootinfo.h
> +++ b/arch/loongarch/include/asm/bootinfo.h
> @@ -28,7 +28,7 @@ struct loongson_board_info {
>  struct loongson_system_configuration {
>         int nr_cpus;
>         int nr_nodes;
> -       int nr_io_pics;
> +//     int nr_io_pics;
>         int boot_cpu_id;
>         int cores_per_node;
>         int cores_per_package;
>
> and see that the kernel still compiles.
>
>         M.
>
> --
> Without deviation from the norm, progress is not possible.