[RFC PATCH 0/7] Pseudo-NMI for arm64 using ICC_PMR_EL1 (GICv3)

[Daniel Thompson]

On 01/04/15 16:29, Marc Zyngier wrote:
> On 01/04/15 16:15, Dave P Martin wrote:
>> Apologies for the slow reply... :/
>>
>> Anyway,
>>
>> On Mon, Mar 23, 2015 at 06:47:53PM +0000, Daniel Thompson wrote:
>>> On 20/03/15 15:45, Dave Martin wrote:
>>>> On Wed, Mar 18, 2015 at 02:20:21PM +0000, Daniel Thompson wrote:
>>>>> This patchset provides a pseudo-NMI for arm64 kernels by reimplementing
>>>>> the irqflags macros to modify the GIC PMR (the priority mask register is
>>>>> accessible as a system register on GICv3 and later) rather than the
>>>>> PSR. The pseudo-NMI changes are support by a prototype implementation of
>>>>> arch_trigger_all_cpu_backtrace that allows the new code to be exercised.
>>
>> Minor nit: the "pseudo NMI" terminology could lead to confusion if
>> something more closely resembling a real NMI comes along.
>>
>> I'll have to have a think, but nothing comes to mind right now...
>>
>> [...]
>>
>>>>>   3. Requires GICv3+ hardware together with firmware support to enable
>>>>>      GICv3 features at EL3. If CONFIG_USE_ICC_SYSREGS_FOR_IRQFLAGS is
>>>>>      enabled the kernel will not boot on older hardware. It will be hard
>>>>>      to diagnose because we will crash very early in the boot (i.e.
>>>>>      before the call to start_kernel). Auto-detection might be possible
>>>>>      but the performance and code size cost of adding conditional code to
>>>>>      the irqflags macros probably makes it impractical. As such it may
>>>>>      never be possible to remove this limitation (although it might be
>>>>>      possible to find a way to survive long enough to panic and show the
>>>>>      results on the console).
>>>>
>>>> This can (and should) be done via patching -- otherwise we risk breaking
>>>> single kernel image for GICv2+v3.
>>>
>>> Do you mean real patching (hunting down all those inlines and
>>> rewrite them) or simply implementing irqflags with an ops table? If
>>> the former I didn't look at this because I didn't release we could
>>> do that...
>>
>> A generic patching framework was introduced by Andre Przywara in this
>> patch:
>>
>> e039ee4 arm64: add alternative runtime patching
>>
>> I believe you should be able to use this to patch between DAIF and
>> ICC_PMR accesses.
>>
>> You should be able to find examples of this framework being used by
>> grepping.  I've not played with it myself yet.
>
> To follow-up on this, I have a few patches queued that use the runtime
> patching code to deal with GICv3 in KVM:
>
> http://thread.gmane.org/gmane.comp.emulators.kvm.arm.devel/616
>
> The first few patches are already queued for v4.1, and the rest should
> follow shortly after.

Thanks (both).

That's really helpful: links for that sort of thing are not easily 
googleable (things like kpatch floods the results).