Re: [PATCH v10 11/20] x86/VPMU: Interface for setting PMU mode and flags

[Boris Ostrovsky <boris.ostrovsky@oracle.com>]

On 09/19/2014 02:51 AM, Jan Beulich wrote:
>>>> On 18.09.14 at 23:50, <boris.ostrovsky@oracle.com> wrote:
>> On 09/18/2014 12:11 AM, Tian, Kevin wrote:
>>>> From: Boris Ostrovsky [mailto:boris.ostrovsky@oracle.com]
>>>> +{
>>>> +    unsigned i, j, allbutself_num, tasknum, mycpu;
>>>> +    static s_time_t start;
>>>> +    static struct tasklet **sync_task;
>>>> +    struct vcpu *curr_vcpu = current;
>>>> +    static struct vcpu *sync_vcpu;
>>>> +    int ret = 0;
>>>> +
>>>> +    tasknum = allbutself_num = num_online_cpus() - 1;
>>>> +
>>>> +    if ( sync_task ) /* if set, we are in hypercall continuation */
>>>> +    {
>>>> +        if ( (sync_vcpu != NULL) && (sync_vcpu != curr_vcpu) )
>>>> +            /* We are not the original caller */
>>>> +            return -EAGAIN;
>>> I assume hypercall continuation will happen on original vcpu context. Under
>>> which situation the hypercall will be continued on a different vcpu? If yes,
>>> would it be an unbounded situation where you may wait unexpected time
>>> to have sync_vcpu==curr_vcpu?
>> The continuation call is gone in v11. If we are stuck waiting for more
>> than 5 seconds we'll simply return -EAGAIN and have the caller retry.
> That reads wrong, but I'd of course have to see whether the just
> wording above is incorrectly reflecting the code changes you did.

The only thing that I can think of that I didn't mention is 
hypercall_preempt_check() to bail out earlier than 5 seconds. Is that 
what you are missing in my reply? Otherwise  I don't know what else is 
wrong.

-boris