Re: [PATCH v2 07/17] irqchip/gic: Atomically update affinity
From: Magnus Damm
Date: Tue Sep 14 2021 - 23:29:16 EST
Hi Geert, everyone,
On Mon, Sep 13, 2021 at 5:05 PM Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:
> Hi Magnus,
> On Sun, Sep 12, 2021 at 7:40 AM Magnus Damm <magnus.damm@xxxxxxxxx> wrote:
> > On Sun, Sep 12, 2021 at 4:32 AM Marc Zyngier <maz@xxxxxxxxxx> wrote:
> > > On Sat, 11 Sep 2021 03:49:20 +0100,
> > > Magnus Damm <magnus.damm@xxxxxxxxx> wrote:
> > > > On Fri, Sep 10, 2021 at 10:19 PM Geert Uytterhoeven
> > > > <geert@xxxxxxxxxxxxxx> wrote:
> > > > > On Fri, Sep 10, 2021 at 12:23 PM Marc Zyngier <maz@xxxxxxxxxx> wrote:
> > > > > > On Thu, 09 Sep 2021 16:22:01 +0100,
> > > > > > Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:
> > > > > GIC: enabling workaround for broken byte access
> > > >
> > > > Indeed, byte access is unsupported according to the EMEV2 documentation.
> > > >
> > > > The EMEV2 documentation R19UH0036EJ0600 Chapter 7 Interrupt Control on
> > > > page 97 says:
> > > > "Interrupt registers can be accessed via the APB bus, in 32-bit units"
> > > > "For details about register functions, see ARM Generic Interrupt
> > > > Controller Architecture Specification Architecture version 1.0"
> > > > The file "R19UH0036EJ0600_1Chip.pdf" is the 6th edition version
> > > > published in 2010 and is not marked as confidential.
> > >
> > > This is as bad as it gets. Do you know if any other Renesas platform
> > > is affected by the same issue?
> > Next time we have a beer together I would be happy to show you some
> > legacy interrupt controller code. =)
> > EMEV2 and the Emma Mobile product line came from the NEC Electronics
> > side that got merged into Renesas Electronics in 2010. Historically
> > NEC Electronics mainly used MIPS I've been told, and the Emma Mobile
> > SoCs were one of the earlier Cortex-A9 adopters. That might have
> > something to do with the rather loose interpretation of the spec.
> Indeed. I used to work on products using EMMA1 and EMMA2, and they
> were MIPS-based (vr4120A for EMMA2, IIRC). Later variants (EMMA2H
> and EMMA3?) did include a small ARM core for standby control.
Thanks for sharing some more background!
> > Renesas SoCs from a similar era:
> > AP4 (sh7372) AP4EVB (Cortex-A8 + INTCA/INTCS)
> This is no longer supported upstream (and not affected, as no GIC).
Right. I might mix it up with the AP4.5 chip that I used for SMP
prototyping back then. It had 4 x CA9 and obviously a GIC.
> > R-Mobile A1 (r8a7740) Armadillo-800-EVA (Cortex-A9 + INTCA/INTCS)
> R-Mobile A1 has GIC (PL390), too, and is not affected.
> > R-Car M1A (r8a7778) Bock-W (Cortex-A9 + GIC)
> > R-Car H1 (r8a7779) Marzen (4 x Cortex-A9 + GIC)
> > Emma Mobile EMEV2 KZM9D (2 x Cortex-A9 + GIC)
> > SH-Mobile AG5 (sh73a0) KZM9G (2 x Cortex-A9 + GIC)
> All of these (except for EMEV2) are fine, too.
Thanks for checking!