2020-03-11 20:18:29

by Tim Harvey

[permalink] [raw]
Subject: CN80xx (octeontx/thunderx) breakage from f2d8340

Marc,

Im seeing a failure to boot on an octeontx CN80xx (thunderx) due to
f2d8340 ("irqchip/gic-v3: Add GICv4.1 VPEID size discovery"). I'm not
sure if something is hanging, I just get no console output from the
kernel.

Is there perhaps something in the dt that requires change? The
board/dts I'm using is:
https://github.com/Gateworks/dts-newport/blob/sdk-10.1.1.0-newport/gw6404-linux.dts
https://github.com/Gateworks/dts-newport/blob/sdk-10.1.1.0-newport/gw640x-linux.dtsi
https://github.com/Gateworks/dts-newport/blob/sdk-10.1.1.0-newport/cn81xx-linux.dtsi

Any ideas? I've cc'd the Cavium/Marvell folk to see if they know
what's up or can reproduce on some of their hardware.

Best Regards,

Tim


2020-03-12 08:56:29

by Marc Zyngier

[permalink] [raw]
Subject: Re: CN80xx (octeontx/thunderx) breakage from f2d8340

Hi Tim,

On 2020-03-11 20:17, Tim Harvey wrote:
> Marc,
>
> Im seeing a failure to boot on an octeontx CN80xx (thunderx) due to
> f2d8340 ("irqchip/gic-v3: Add GICv4.1 VPEID size discovery"). I'm not
> sure if something is hanging, I just get no console output from the
> kernel.

That's odd. It probably means that a SError has been taken to EL3,
and the firmware is not equipped to deal with it. Great stuff!

> Is there perhaps something in the dt that requires change? The
> board/dts I'm using is:
> https://github.com/Gateworks/dts-newport/blob/sdk-10.1.1.0-newport/gw6404-linux.dts
> https://github.com/Gateworks/dts-newport/blob/sdk-10.1.1.0-newport/gw640x-linux.dtsi
> https://github.com/Gateworks/dts-newport/blob/sdk-10.1.1.0-newport/cn81xx-linux.dtsi
>
> Any ideas? I've cc'd the Cavium/Marvell folk to see if they know
> what's up or can reproduce on some of their hardware.

This is most probably Cavium erratum 38539. Please give [1] a go and
let me know whether it helps by replying to the patch.

Thanks,

M.

[1] https://lore.kernel.org/lkml/[email protected]/
--
Jazz is not dead. It just smells funny...

2020-03-12 09:45:39

by Robert Richter

[permalink] [raw]
Subject: Re: CN80xx (octeontx/thunderx) breakage from f2d8340

Adding Sunil here.

-Robert

On 12.03.20 08:55:52, Marc Zyngier wrote:
> Hi Tim,
>
> On 2020-03-11 20:17, Tim Harvey wrote:
> > Marc,
> >
> > Im seeing a failure to boot on an octeontx CN80xx (thunderx) due to
> > f2d8340 ("irqchip/gic-v3: Add GICv4.1 VPEID size discovery"). I'm not
> > sure if something is hanging, I just get no console output from the
> > kernel.
>
> That's odd. It probably means that a SError has been taken to EL3,
> and the firmware is not equipped to deal with it. Great stuff!
>
> > Is there perhaps something in the dt that requires change? The
> > board/dts I'm using is:
> > https://github.com/Gateworks/dts-newport/blob/sdk-10.1.1.0-newport/gw6404-linux.dts https://github.com/Gateworks/dts-newport/blob/sdk-10.1.1.0-newport/gw640x-linux.dtsi https://github.com/Gateworks/dts-newport/blob/sdk-10.1.1.0-newport/cn81xx-linux.dtsi
> >
> > Any ideas? I've cc'd the Cavium/Marvell folk to see if they know
> > what's up or can reproduce on some of their hardware.
>
> This is most probably Cavium erratum 38539. Please give [1] a go and
> let me know whether it helps by replying to the patch.
>
> Thanks,
>
> M.
>
> [1] https://lore.kernel.org/lkml/[email protected]/
> --
> Jazz is not dead. It just smells funny...

2020-03-12 09:59:17

by Robert Richter

[permalink] [raw]
Subject: Re: CN80xx (octeontx/thunderx) breakage from f2d8340

Tim,

> > On 2020-03-11 20:17, Tim Harvey wrote:

> > > Im seeing a failure to boot on an octeontx CN80xx (thunderx) due to
> > > f2d8340 ("irqchip/gic-v3: Add GICv4.1 VPEID size discovery"). I'm not
> > > sure if something is hanging, I just get no console output from the
> > > kernel.

if you are booting with serial console, you could try to use earlycon
boot parameter to get more output. Though, I do not have the
parameters for that system at hand. You could try 'earlycon' without
parameters or perhaps 'earlycon=pl011,0x87e024000000' or similar (not
sure if that works, maybe check devicetree).

-Robert

2020-03-12 16:59:49

by Tim Harvey

[permalink] [raw]
Subject: Re: CN80xx (octeontx/thunderx) breakage from f2d8340

On Thu, Mar 12, 2020 at 1:55 AM Marc Zyngier <[email protected]> wrote:
>
> Hi Tim,
>
> On 2020-03-11 20:17, Tim Harvey wrote:
> > Marc,
> >
> > Im seeing a failure to boot on an octeontx CN80xx (thunderx) due to
> > f2d8340 ("irqchip/gic-v3: Add GICv4.1 VPEID size discovery"). I'm not
> > sure if something is hanging, I just get no console output from the
> > kernel.
>
> That's odd. It probably means that a SError has been taken to EL3,
> and the firmware is not equipped to deal with it. Great stuff!
>
> > Is there perhaps something in the dt that requires change? The
> > board/dts I'm using is:
> > https://github.com/Gateworks/dts-newport/blob/sdk-10.1.1.0-newport/gw6404-linux.dts
> > https://github.com/Gateworks/dts-newport/blob/sdk-10.1.1.0-newport/gw640x-linux.dtsi
> > https://github.com/Gateworks/dts-newport/blob/sdk-10.1.1.0-newport/cn81xx-linux.dtsi
> >
> > Any ideas? I've cc'd the Cavium/Marvell folk to see if they know
> > what's up or can reproduce on some of their hardware.
>
> This is most probably Cavium erratum 38539. Please give [1] a go and
> let me know whether it helps by replying to the patch.
>
>
> [1] https://lore.kernel.org/lkml/[email protected]/

Marc,

Yup, this was it! We need to make sure this gets merged into 5.6. I
didn't have the original patch but attempted to reply via mailto:
link.

Best regards,

Tim