Re: [PATCH v3 00/16] TDX: Enable Dynamic PAMT

[Dave Hansen]

On 9/26/25 09:02, Edgecombe, Rick P wrote:
> On Fri, 2025-09-26 at 07:09 -0700, Dave Hansen wrote:
>> On 9/25/25 19:28, Yan Zhao wrote:
>>>> Lastly, Yan raised some last minute doubts internally about TDX
>>>> module locking contention. I’m not sure there is a problem, but
>>>> we
>>>> can come to an agreement as part of the review.
>>> Yes, I found a contention issue that prevents us from dropping the
>>> global lock. I've also written a sample test that demonstrates this
>>> contention.
>> But what is the end result when this contention happens? Does
>> everyone livelock?
> 
> You get a TDX_OPERAND_BUSY error code returned. Inside the TDX module
> each lock is a try lock. The TDX module tries to take a sequence of
> locks and if it meets any contention it will release them all and
> return the TDX_OPERAND_BUSY. Some paths in KVM cannot handle failure
> and so don't have a way to handle the error.
> 
> So another option to handling this is just to retry until you succeed.
> Then you have a very strange spinlock with a heavyweight inner loop.
> But since each time the locks are released, some astronomical bad
> timing might prevent forward progress. On the KVM side we have avoided
> looping. Although, I think we have not exhausted this path.

If it can't return failure then the _only_ other option is to spin. Right?

I understand the reluctance to have such a nasty spin loop. But other
than reworking the KVM code to do the retries at a higher level, is
there another option?