[PATCH v4 00/21] SError rework + RAS&IESB for firmware first support

[James Morse]

Hi Drew,

On 13/11/17 16:14, Andrew Jones wrote:
> On Mon, Nov 13, 2017 at 12:29:46PM +0100, Christoffer Dall wrote:
>> On Thu, Nov 09, 2017 at 06:14:56PM +0000, James Morse wrote:
>>> On 19/10/17 15:57, James Morse wrote:
>>>> Known issues:
>>>>  * KVM-Migration: VDISR_EL2 is exposed to userspace as DISR_EL1, but how should
>>>>    HCR_EL2.VSE or VSESR_EL2 be migrated when the guest has an SError pending but
>>>>    hasn't taken it yet...?
>>>
>>> I've been trying to work out how this pending-SError-migration could work.

[..]

>>> To get out of this corner: why not declare pending-SError-migration an invalid
>>> thing to do?
>>
>> To answer that question we'd have to know if that is generally a valid
>> thing to require.  How will higher level tools in the stack deal with
>> this (e.g. libvirt, and OpenStack).  Is it really valid to tell them
>> "nope, can't migrate right now".  I'm thinking if you have a failing
>> host and want to signal some error to the guest, that's probably a
>> really good time to migrate your mission-critical VM away to a different
>> host, and being told, "sorry, cannot do this" would be painful.  I'm
>> cc'ing Drew for his insight into libvirt and how this is done on x86,
>> but I'm not really crazy about this idea.

> Without actually confirming, I'm pretty sure it's handled with a best
> effort to cancel the migration, continuing/restoring execution on the
> source host (or there may be other policies that could be set as well).
> Naturally, if the source host is going down and the migration is
> cancelled, then the VM goes down too...

> Anyway, I don't think we would generally want to introduce guest
> controlled migration blockers. IIUC, this migration blocker would remain
> until the guest handled the SError, which it may never unmask.

Yes, given the guest can influence this it needs exposing so it can be migrated.


[...]

>> My suggestion would be to add some set of VCPU exception state,
>> potentially as flags, which can be migrated along with the VM, or at
>> least used by userspace to query the state of the VM, if there exists a
>> reliable mechanism to restore the state again without any side effects.
>>
>> I think we have to comb through Documentation/virtual/kvm/api.txt to see
>> if we can reuse anything, and if not, add something.  We could also
> 
> Maybe KVM_GET/SET_VCPU_EVENTS? Looks like the doc mistakenly states it's
> a VM ioctl, but it's a VCPU ioctl.

Hmm, if I suppress my register-size pedantry we can put the lower 32 bits of
VSESR_EL2 in exception.error_code and use has_error_code to mark it valid.
'exception' in this struct ends up meaning SError on arm64.

(While VSESR_EL2 is 64bit[0], the value gets written into the ESR, which is
32bit, so I doubt the top 32bits can be used, currently they are all reserved.)

I'll go dig into how x86 uses this...


Thanks!

James


[0]
https://static.docs.arm.com/ddi0587/a/RAS%20Extension-release%20candidate_march_29.pdf