Warning on SME when manually onlining CPUs late.

Warning on SME when manually onlining CPUs late.

[Jonathan Cameron]

Hi Mark

Hit this due to a silly config error on a QEMU TCG test but seemed worth asking about
as I'm buried in other stuff and no time to investigate further for a few weeks.

Test probably could be simplified but it's more or less vanilla upstream QEMU with
-cpu max -smp 8 and kernel command line with maxcpus=4

echo 1 > /sys/bus/cpu/devices/cpu4/online

CPU features: SANITY CHECK: Unexpected variation in SYS_SMCR_EL1. Boot CPU: 0x0000000000000f, CPU4: 0x0000008000000f
CPU features: Unsupported CPU feature variation detected.

If this is already well known, then sorry for noise. I've not been paying much attention
to the SME work.

Looks like the FA64 bit isn't set for the CPU state of the boot CPU that is being compared
- I guess a stale record somewhere being used for the comparison?

Could be an issue on QEMU side I guess - though from a quick look it all seemed to be
correct there.  Just in case +CC Richard.

Jonathan

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: Warning on SME when manually onlining CPUs late.

[Richard Henderson]

On 7/27/23 03:21, Jonathan Cameron wrote:
> Hi Mark
> 
> Hit this due to a silly config error on a QEMU TCG test but seemed worth asking about
> as I'm buried in other stuff and no time to investigate further for a few weeks.
> 
> Test probably could be simplified but it's more or less vanilla upstream QEMU with
> -cpu max -smp 8 and kernel command line with maxcpus=4
> 
> echo 1 > /sys/bus/cpu/devices/cpu4/online
> 
> CPU features: SANITY CHECK: Unexpected variation in SYS_SMCR_EL1. Boot CPU: 0x0000000000000f, CPU4: 0x0000008000000f
> CPU features: Unsupported CPU feature variation detected.
> 
> If this is already well known, then sorry for noise. I've not been paying much attention
> to the SME work.
> 
> Looks like the FA64 bit isn't set for the CPU state of the boot CPU that is being compared
> - I guess a stale record somewhere being used for the comparison?
> 
> Could be an issue on QEMU side I guess - though from a quick look it all seemed to be
> correct there.  Just in case +CC Richard.

SMCR_EL1 is for the OS to configure -- if they have different values then that's a kernel 
bug.  I wouldn't expect this to be a QEMU problem.


r~


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: Warning on SME when manually onlining CPUs late.

[Mark Brown]

[-- Attachment #1.1: Type: text/plain, Size: 271 bytes --]

On Thu, Jul 27, 2023 at 10:02:24AM -0700, Richard Henderson wrote:

> SMCR_EL1 is for the OS to configure -- if they have different values then
> that's a kernel bug.  I wouldn't expect this to be a QEMU problem.

It's not, I have a patch that should go out later today.

[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

[-- Attachment #2: Type: text/plain, Size: 176 bytes --]

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel