[PATCH v2] KVM: arm/arm64: Signal SIGBUS when stage2 discovers hwpoison memory

[Peter Maydell]

On 23 June 2017 at 10:38, James Morse <james.morse at arm.com> wrote:
> So the v8 ARM-ARM has two ways of indicating an external abort, KVM tests bit 9
> The v7 ARM-ARm has a '{D,I}FSR.ExT' bit, which I think is equivalent. Its
> described as:
>> An implementation can use the DFSR.ExT and IFSR.ExT bits to provide more
>> information about external aborts.

For instance, the Cortex-A15 TRM documents that it uses the ExT bit
to distinguish AXI DECERR (decode error, ExT==0) and SLVERR (slave
error, ExT==1) abort responses.

>> This is what the v8 version must mean with its
>> External abort type. This bit can provide an IMPLEMENTATION DEFINED
>> classification of External aborts.
>
> Which I read as IMP-DEF classification 'as external', as opposed to your reading
> as an extra IMP-DEF classification for external aborts.

I think gengdongjiu's reading is correct here: if this
is an external abort then you can find out more info about
exactly what kind of EA it is by looking at this bit, but
you have to look at the FSC bits to be sure it is an EA first.

The Cortex-A53 TRM documents that it uses the same SLVERR-vs-DECERR
semantics for DFSR.ExT as the A15. It doesn't say so but it wouldn't
be very surprising if its usage of ESR_ELx.EA was the same.

thanks
-- PMM